SONAR - Dithering

SONAR - Dithering
Tech support informed me to check the manual (and I did)
 
Apparently I need the idiot version:  
 
Rectangular. Essentially white noise, no noise shaping. Advantages: least CPU-intensive, lowest signal-to-noise ratio, preferable to shaped dither when successive dithering can occur (e.g. bouncing, freezing). Disadvantages: suffers from intermodulation distortion, higher perceived loudness than Pow-r dither.

Triangular. Higher level than rectangular, no noise shaping. Advantages: low CPU-intensive dither, superior to Rectangular as it does not suffer from modulation noise effects. Preferable to shaped (Pow-r) dither when successive dithering can occur (e.g. bouncing, freezing). Disadvantages: higher perceived loudness than Pow-r dither.

Pow-r 1. Noise-shaped dither. Advantages: less CPU-intensive than Pow-r types 2 and 3, lower perceived loudness than Rectangular or Triangular. Disadvantages: less noise shaping than Pow-r types 2 and 3, not recommended for operations where dither will be applied successively (e.g. bounce and freeze).

Pow-r 2. Noise-shaped dither. Advantages: lowest perceived loudness, highest quality settings, recommended for audio export. Disadvantages: highest CPU-intensive settings, not recommended for operations where dither will be applied successively (e.g. bounce and freeze).

Pow-r 3. Same as Pow-r 2 except most CPU-intensive and transparent of all choices.

 
What I'll be recording Example:
Sampling Rate: 48 24bit Depth
Distorded electric guitar - Direct (no open mic)
Classical (Nylon Strings) - Direct & Open Mic
Acoustic (Steel Strings) - Direct & Open Mic
Vocals - Open mic (direct possible when the Matrix takes over)
 
What, (if any)changes should I make from the default Triangular?
 

post edited by sadicus - September 20, 09 11:17 AM

Turn all dithering off in Sonar and leave it that way. Export your "masters" at 24-bit depth. Get Voxengo's *free* R8brain utility: http://www.voxengo.com/product/r8brain/ and use that to convert your 24-bit masters to 16-bit CD tracks. R8brain Pro http://www.voxengo.com/product/r8brainpro/ is even better, tho not free. Either will do a significantly better job than Sonar at converting the word-depth - esp clarity in the extreme high-end. Once I started using R8brain, I never used Sonar dithering again.
-Dan

wintaper


Turn all dithering off in Sonar and leave it that way. Export your "masters" at 24-bit depth. Get Voxengo's *free* R8brain utility: http://www.voxengo.com/product/r8brain/ and use that to convert your 24-bit masters to 16-bit CD tracks. R8brain Pro http://www.voxengo.com/product/r8brainpro/ is even better, tho not free. Either will do a significantly better job than Sonar at converting the word-depth - esp clarity in the extreme high-end. Once I started using R8brain, I never used Sonar dithering again.
-Dan

I have Sound Forge 8d-would the R8Brain be better, the same or not as good? Dithering has always caused me some indecision
Thanks
Frank

stratman70

Dithering has always caused me some indecision

Well, first of all don't lose sleep over it. Despite what people might say, the differences are fairly subtle (remember that when it comes to audio, people tend to think every difference is either inaudible or a huge/night and day difference - there seems to be no middle ground).

Dither is just noise added to the very quietest part of a signal to try to preserve some of the data there that otherwise gets lost when you reduce the number of bits. "Noise shaped" dither tries to put most of the noise at (very high) frequencies where we won't notice it as much.

As stated, just make sure that dither is the very last thing you do when reducing bits. You don't want to add noise to your audio, then perform other tasks that might magnify the noise.

drewfx

OK-Guys. I am bouncing softsynths at 32bit (which sounds nice) so I do have to dither on final (for CD).

No you dont need to dither going up. Can you, yes. should you, No. Only dither at your very last step of your project, form 24bit (or 32bit) to 16bit.
Cj

Remember also that you can audibly test dither differences by mixing two files with different dither settings, synching them in Sonar with one file's phase inverted. If they're the same, the phase invertion cancels the other file out and you hear nothing. If they're not the same, whatever you would hear would be the difference that one method has vs. the other.

CJaysMusic

OK-Guys. I am bouncing softsynths at 32bit (which sounds nice) so I do have to dither on final (for CD).

No you dont need to dither going up. Can you, yes. should you, No. Only dither at your very last step of your project, form 24bit (or 32bit) to 16bit.
Cj

That's what I meant, I do have to dither on final (for CD ) because the tracks are 32bit. I didn't word it quite right :)  But I am still curious-I have sound forge 8d, but you guys are saying I shouldn't use that to dither-why not? I have read here many times that SF8 was used by many. I did download R8Brain -free. Has anyone compared Sonar 85 powr-3 - to Sound forge 8d to R8Brain?
Thanks for all the help-Love this forum!

In addition to what everyone else in this thread has already said, be aware of the following:

1) Some experienced audio professionals and programmers indicate that it is a good idea to dither anytime the results of audio mathematical calculations get fixed into permanent form - in other words, anytime you bounce anything (whether it's the result of submixing tracks to free up resources, or destructively applying an audio effect, etc.), you should be dithering to avoid introducing quantization errors that could be later magnified through further mixing or mastering processes.  However, all "intermediate" dithering (i.e., *everything* except dithering on the final master) should be non-noise-shaping (NOT the Pow-R algorithms).

2) Some experienced audio professionals and programmers also indicate that if you *do* dither, you should ONLY use noise-shaping dither as the LAST dither in the chain.  In other words, if your final mastering process, that created the 16-bit files that will actually go on a CD, will include dithering, *that's* where you should be using a noise-shaping dither like the Pow-R algorithms.  Any other dither at any stage should be non-noise-shaping, like triangular.  This is because any further mixing or mastering processes - including the compression and limiting that are typically part of mastering - can magnify the dither noise that noise-shaping dithers deliberately introduce into the audio spectrum.

CJaysMusic

But I am still curious-I have sound forge 8d, but you guys are saying I shouldn't use that to dither-why not? I have read here many times that SF8 was used by many. I did download R8Brain -free. Has anyone compared Sonar 85 powr-3 - to Sound forge 8d to R8Brain? Thanks for all the help-Love this forum!

uoi guys said?? I never said not to use the dither in sound forge. If you like it better thabn the POw3 or POW2 or Pow1 ,then use it, but only dither at the very last step of your process, when your exporting to 44khz/16bit.  Is that clear enough. what dither you use is up to you
Cj

Woah-I didn't say you said anything.  Sorry! I asked about dithering with SF and the other guys (OK the other guys) not CJ start talking about using R8Brain and powr-3. And if you read my 2nd post you will see that I said I would dither when I master to cd-as in Final. I just am trying to get info from the people I feel know much more than I do about this (that includes you also, I have gotten many great tips and hints from your posts over the years)

rstollen


stratman - I'm with you on dithering with Sound Forge. You would export from Sonar at 24-bits and no dithering. Then bring the exported file into Sound Forge to perform a bit reduction with dither.

OK Thanks-Things are a bit (no pun) different for me I think because only guitars and vocals are recorded live. Drums, bass and keys are all midi using softsynths. So I bounce the softsynths at 32bit (default, I never touched it. But I always left the dither option on (when bouncing the ss) defaulted to triangular? So should I change the render bounce setting  (global options) to 24Bits instead of 32bits? Wish I could use Sonar as well as I can play my guitars. But I cannot and I imagine I never will. But I sure want to get better. Plus the difference/actually relationship) between "audio driver bit depth (mine is set at 16bit-it's capable of 24/96) vs recording bit depth (which is at 24 bits)? Mind boggling for this guiatr palyer

RTGraham


In addition to what everyone else in this thread has already said, be aware of the following:

1) Some experienced audio professionals and programmers indicate that it is a good idea to dither anytime the results of audio mathematical calculations get fixed into permanent form - in other words, anytime you bounce anything (whether it's the result of submixing tracks to free up resources, or destructively applying an audio effect, etc.), you should be dithering to avoid introducing quantization errors that could be later magnified through further mixing or mastering processes.  However, all "intermediate" dithering (i.e., *everything* except dithering on the final master) should be non-noise-shaping (NOT the Pow-R algorithms).

2) Some experienced audio professionals and programmers also indicate that if you *do* dither, you should ONLY use noise-shaping dither as the LAST dither in the chain.  In other words, if your final mastering process, that created the 16-bit files that will actually go on a CD, will include dithering, *that's* where you should be using a noise-shaping dither like the Pow-R algorithms.  Any other dither at any stage should be non-noise-shaping, like triangular.  This is because any further mixing or mastering processes - including the compression and limiting that are typically part of mastering - can magnify the dither noise that noise-shaping dithers deliberately introduce into the audio spectrum.

+1  
I agree with everything Graham said here.  IOW, DO NOT turn Dithering OFF in Sonar.
 
Leave "Triangular" dithering ON.  This is a non-noise shaping dithering algorithm that's used, not only when bouncing/freezing files to 24 bit internally in Sonar, but also when sound comes out through your audio interface.  Remember that audio interfaces operate at 24 bits which means that Sonar's 32/64 bit processing is being truncated on output unless you have Triangular dithering ON (under Options::Audio::Advanced Tab).  As Graham said, truncation adds quantization errors to your music which is what dithering avoids.  Maybe you won't hear it, but it doesn't mean it's not there!
 
Also, Sonar does NOT dither when the files are kept in floating point format (32 or 64 bits).  It only applies dithering when going down to either 24 or 16 bits (which is the correct time to dither).  This means that, even if you have Triangular dithering enabled, your files will ONLY get dithered when they need to be.  Sonar handles this automatically for you, so you don't have to worry about it.  All you have to do is apply noise-shaping dithering (like POW-R or MBIT+ from Izotope) as the final step in Mastering.  That's is all!
 
So leave the dithering option enabled (it was the default setting for a reason), and only worry about it at the end (in the Mastering process).
 
 
HTH

stratman70

rstollen


stratman - I'm with you on dithering with Sound Forge. You would export from Sonar at 24-bits and no dithering. Then bring the exported file into Sound Forge to perform a bit reduction with dither.

OK Thanks-Things are a bit (no pun) different for me I think because only guitars and vocals are recorded live. Drums, bass and keys are all midi using softsynths. So I bounce the softsynths at 32bit (default, I never touched it. But I always left the dither option on (when bouncing the ss) defaulted to triangular? So should I change the render bounce setting  (global options) to 24Bits instead of 32bits? Wish I could use Sonar as well as I can play my guitars. But I cannot and I imagine I never will. But I sure want to get better. Plus the difference/actually relationship) between "audio driver bit depth (mine is set at 16bit-it's capable of 24/96) vs recording bit depth (which is at 24 bits)? Mind boggling for this guiatr palyer

I don't want to pretend to be an expert on whether you should dither on a bounce. I was assuming that your project was at 24-bits, and that you would leave it at that setting when exporting. 

stratman70

So should I change the render bounce setting  (global options) to 24Bits instead of 32bits? Wish I could use Sonar as well as I can play my guitars. But I cannot and I imagine I never will. But I sure want to get better. Plus the difference/actually relationship) between "audio driver bit depth (mine is set at 16bit-it's capable of 24/96) vs recording bit depth (which is at 24 bits)? Mind boggling for this guiatr palyer

I'd leave the bounce settings the way they came in Sonar (at 32 bits).  That way you, not only keep things in floating point format (avoiding either truncation if dithering was turned OFF, or the actual dithering process if it's ON), but also keep the wide dynamic range and precision of the 32 bit floating point file format.
 
The 'Audio Driver Bit Depth' pertains to the bitness of what you hear coming out of your speakers.  The 'Record Bit Depth', on the other hand, is the bit depth of your recordings.
 
 
HTH

The fact that we have to dither at all is just an aspect of the imperfections that exist in digital music and the formats we are constrained to use to reach the "market". There is no dithering program that preserves the original file. 48-44.1, 24 bit-16 bit. Once you start taking out peices of the data that an algorythm has control over, unless you have conformed to that algorythm while recording, you may lose some important data.
Should we all record in 16 bit to begin with to avoid dithering? When we convert to Mp3 does it recognize even the 16 bit file accurately? It's all part of a fight between entities in music reproduction to control the market and we are caught in the frey. Progress is slow, new mediums are in flux therefore. How can you trust your ears when you compose in a better format and then are forced to reduce to a lessor format. With dithering, I can't get no satisfaction. 

Jose7822

stratman70

So should I change the render bounce setting  (global options) to 24Bits instead of 32bits? Wish I could use Sonar as well as I can play my guitars. But I cannot and I imagine I never will. But I sure want to get better. Plus the difference/actually relationship) between "audio driver bit depth (mine is set at 16bit-it's capable of 24/96) vs recording bit depth (which is at 24 bits)? Mind boggling for this guiatr palyer

I'd leave the bounce settings the way they came in Sonar (at 32 bits).  That way you, not only keep things in floating point format (avoiding either truncation if dithering was turned OFF, or the actual dithering process if it's ON), but also keep the wide dynamic range and precision of the 32 bit floating point file format.
 
The 'Audio Driver Bit Depth' pertains to the bitness of what you hear coming out of your speakers.  The 'Record Bit Depth', on the other hand, is the bit depth of your recordings.
 
 
HTH

Thanks Jose - That's one of the things I was looking for!
and rstollen- I will use Sf8d, I mean I do own it and it works well. so thanks
Great info all around.